Building a DIY Music Server

[littlej0e]

1 minute ago, JackJohnson316 said:

Will the Taiko USB card work in a DIY Extreme clone? If so I'm interested too. I thought I read Taiko had to make custom drivers for every dac, so if my dac isn't on the driver list I'm not sure it would work.

 

Yes, the card itself should work fine. However, the drivers seem to be the main sticking point. I believe they are highly customized and this is the primary reason why Taiko doesn't want to release them or sell the USB card to non Extreme owners. I don't necessarily blame them, but it won't stop me from begging incessantly haha. 

 

I've been pining, searching, begging and emailing and since last July with zero luck:-(

[Kuba7777]

35 minutes ago, littlej0e said:

Chciałbym też dołożyć @Kuba7777prośba o konfiguracje ProcessLasso. Zdaję sobie sprawę, że dla wielu jest to terytorium „tajnego sosu”, ale dzielenie się oznacza troskę! 

thank you very much for your answer. Yesterday I configured ProcessLasso almost exactly like you. I assigned everything I could regarding the Windows system and the ProcessLasso program to CPU 1. However, some processes, such as: service.exe, system.exe, registry, csrss.exe, svchost.exe, wininit.exe, I could not assign only to CPU 1. Roon, HQPleyer and all related services to CPU 2. I just don't know how to assign the ASIO driver?

[littlej0e]

34 minutes ago, Kuba7777 said:

thank you very much for your answer. Yesterday I configured ProcessLasso almost exactly like you. I assigned everything I could regarding the Windows system and the ProcessLasso program to CPU 1. However, some processes, such as: service.exe, system.exe, registry, csrss.exe, svchost.exe, wininit.exe, I could not assign only to CPU 1. Roon, HQPleyer and all related services to CPU 2. I just don't know how to assign the ASIO driver?

 

You are quite welcome and thanks for sharing your configs as well!

 

The ASIO driver is usually a stand-alone piece of software that runs like any other application, though Windows doesn't typically come with a driver pre-installed. You can get an open-source driver here: https://asio4all.org/ but I would HIGHLY recommend downloading and installing the purpose-built driver directly from your DAC manufacturer. In my case, I'm running a T+A DAC 200, so this is what my ASIO configuration looks like (see attached). After you've downloaded and installed your ASIO driver, assign the CPU I/O, Mem, etc. values like any other program - done!

 

[littlej0e]

I've been running HQPlayer quite a bit on my DIY rig, so I thought I would share some of the things I learned along the way:

• HQPlayer (Jussi) does not recommend altering the default values, priorities, etc. as this can cause instability, poor performance and adversely affect SQ. Naturally, I maxed out all the settings anyway lol. It's been running quite well thus far.
• I found that running Roon and upsampling together on CPU2 with only 4 or 5 cores assigned to Roon (and the rest assigned to HQPlayer) tended to yield the best sound quality. I haven't' a clue why as I assumed separating these functions on different CPUs would have yielded the best results, but it didn't. At least not on my rig.
• Try using HQPlayer (without upsampling) to play music instead of Roon. Uglier, less feature-rich and a bit more logistically challenging than Roon. But the bump in sound quality might be worth it for some. Turns out HQPlayer is actually a really good music player (with a name like HQPlayer who would have guessed?!?). Now if I could only get my hands on Taiko's XDMS...
• I largely stopped inline upsampling w/HQPlayer because:
  • My DAC (T+A 200) has a built-in native 1 bit DSD DAC that sounds vastly superior to the PCM side. It's not even close. It adds slightly more realism, slightly bigger staging, and far more dynamic rage (especially when playing native DSD files).
  • I discovered dBpoweramp software. I resampled/converted most of my library from 44.1kHz .wave to DSD256 because of the synergy mentioned above. 
    • Running dual Xeon Gold 6148 CPUs on a custom water cooling loop really came in handy. It only took approx. 14 hours to convert approx. 6,000 tracks from 44khz .wav to DSD256 .dif. 
    • Be very careful with this software. It will consume all available system resources and almost melted my rig! 
    • I wouldn't recommend going to all of this trouble unless you have similar native DSD capabilities with a similarly optimized setup. In the end, this is just a "poor man's" way of squeezing better performance out of cheaper gear. Probably isn't worth the hassle for most. 
  • To the surprise of absolutely no one, forgoing upsampling results is lower resource utilization and almost universally better SQ. The same is true when physically decoupling HQPlayer upsampling from the DIY Taiko server as well. 
• Don't forget to configure ProcessLasso to automatically restart applications. For example, open PL then go to Options > Control > then add "c:\program files\signalyst\hqplayer 5 desktop\hqplayer5desktop.exe" to the list then click OK. PL will now auto restart HQPlayer if it crashes, which it can be prone to do. This app/process manager is invaluable for automating your server. 

Hope this helps!

 

- lj 

 

[seeteeyou]

https://audiophilestyle.com/ca/reviews/reality-quest-going-to-extremes-with-the-taiko-audio-sgm-extreme-part-2-of-5-r908/

On 5/19/2020 at 12:56 PM, ray-dude said:

As shown in the block diagram above, by default each PCIe slot has an affinity for one or the other of the CPUs:

• CPU 1 - 1 (16x, blank), 2 (8x, SFP card), 6 (8x, blank), 7 (16x, Music storage card)
• CPU 2 - 3 (8x, OS storage card), 4 (8x, blank), 5 (16x, blank)

This is another tuning opportunity to have I/O cards more directly linked to the CPUs where dependent processes are running. In the case of the Extreme, CPU 1 has direct affinity to all music I/O (network, USB, storage), with CPU 2 delegated to OS I/O. By delegating system and music processes to cores on the appropriate CPU, Emile is able to achieve an optimal balance for sound quality.

 

If I weren't mistaken, simply replicating what Emile already did in the past might be a good place to start?

 

Speaking of PCIe, definitely check this out

 

https://github.com/spddl/GoInterruptPolicy

https://translate.google.com/translate?hl=en&sl=fr&u=http://forum-hifi.fr/thread-18799-page-14.html

http://jplay.eu/forum/index.php?/topic/4410-windows-11-pe-audiophile-creation-guide/?p=67182

Quote

Thank you seeteeyou! A couple of links was so much useful to clear some mess about timers for me. And GoInterruptPolicy app changed so much in sound.

 

[Kuba7777]

5 hours ago, seeteeyou said:

In the case of the Extreme, CPU 1 has direct affinity to all music I/O (network, USB, storage), with CPU 2 delegated to OS I/O. By delegating system and music processes to cores on the appropriate CPU, Emile is able to achieve an optimal balance for sound quality.

 

If I understand correctly, it is better to assign all processes of the Windows operating system and the ProcessLasso program to CPU 2, and services related to audio Roon, HQPlayer, ASIO driver to CPU 1?

[Romaniac]

15 hours ago, Kuba7777 said:

However, some processes, such as: service.exe, system.exe, registry, csrss.exe, svchost.exe, wininit.exe, I could not assign only to CPU 1

To be able to change affinities and priorities of such processes you should use "System Informer" (former "Process Hacker"). It comes with its own kernel driver that allows to manipulate these processes.

[Romaniac]

15 hours ago, littlej0e said:

so this is what my ASIO configuration looks like (see attached)

This is not your ASIO driver. This process is settins panel of your driver. You can easily kill it and listen your music.  

[littlej0e]

7 hours ago, Romaniac said:

This is not your ASIO driver. This process is settins panel of your driver. You can easily kill it and listen your music.  

 

Yes, this is just the front-end application for the ASIO driver running under the OS. I didn't want to pedantically split hairs and confuse the guy. 

[JackJohnson316]

Any recommendations for a case to house the Asus WS C621E Sage motherboard with two Xeon 4210 cpus (Taiko Extreme DIY clone)? I'm looking to buy a temporary case (hopefully cheap) while I wait on HDPLEX to come out with their case that will house the clone. I'm thinking about popping two passive heatsinks on top of the two cpus: 

https://www.newegg.com/dynatron-b12/p/N82E16835114151

 

Thanks for any help.

 

 

[seeteeyou]

System Informer should be considered one of those "secret weapons" that @internethandle explained what could be done about it

 

http://jplay.eu/forum/index.php?/topic/4410-windows-11-pe-audiophile-creation-guide/?p=63761

http://jplay.eu/forum/index.php?/topic/4410-windows-11-pe-audiophile-creation-guide/?p=65646

 

On the same forum thread, similar replies were also posted back in 2021

 

http://jplay.eu/forum/index.php?/topic/4410-windows-11-pe-audiophile-creation-guide/?p=57915

http://jplay.eu/forum/index.php?/topic/4410-windows-11-pe-audiophile-creation-guide/?p=58112

[seeteeyou]

https://www.whatsbestforum.com/threads/introducing-olympus-olympus-i-o-a-new-perspective-on-modern-music-playback.37939/page-21#post-932339

Quote

We’re using a 4th gen AMD Epyc. This is actually a quad CPU centered around an I/O die on a single socket and can be utilised in the same way we do with the dual CPUs of the Extreme.

 

While we only have a limited number of motherboards available for now, we could still aim for a fanless config with either 8 or 16 cores

 

https://servers.asus.com/products/servers/server-motherboards/S14NA-U12

https://www.servethehome.com/tag/siena/

 

[MarcelNL]

Question is why would we?

 

If you read the thread through I read between the lines that Taiko needed to replace the Xeon design as the Silver CPU is no longer available in the quantities they need but that they researched and found this solution to sound quite similar, not better than Sage plus dual Xeon.

 

As Xeon Silver CPU's are pretty sturdy used CPUs should prove value for money for the DIY community IMHO.

 

 

BTW: I seem to have forgotten the magic word to get an NVME in an NVME to PCI adapter to be seen in BIOS, it was a combination of CPU storage setting (not data, so two options left) and formatting to the right boot system and OS...

[lwr]

1 hour ago, MarcelNL said:

BTW: I seem to have forgotten the magic word to get an NVME in an NVME to PCI adapter to be seen in BIOS, it was a combination of CPU storage setting (not data, so two options left) and formatting to the right boot system and OS...

From the SuperMicro installation guide:

[MarcelNL]

1 hour ago, lwr said:

From the SuperMicro installation guide:

Thanks, yet this is the Asus Sage C621E MB, with more settings than I currently seem to be able to handle ;-)

 

There is a bios setting for CPU storage, which has three options, and the formatting settings for the NVME disc in f.e. Rufus provide like 4-5 options for OS, and then the boot system.  

 

Weird thing s that I had both Daphile and W 10 PE working on two NVME SSD drives but they no longer boot...nor are they recognized as NVME under 'NVME settings'

[Iving]

On 4/4/2021 at 9:21 PM, Topk said:

This is another item that I didn’t see much discussion about.

 

On 4/5/2021 at 10:47 AM, MarcelNL said:

as far as I can tell the Elfidelity device is simply a bunch of capacitors stuck on a PCIe circuit board.

 

I predict the effect is small

 

A respectable member here, during the course of a conversation about something else, btw'd me to the effect, "Use a DC filter in the next rail to the pcie card".

 

Google being my friend, and having enquired politely of said Oracle, I could only think the dear fellow was referring to the polychromatic Elfidelity AXF-107.

 

I got one shipped to me from China. The results were noteworthy - and here they are …

 

My system is a PC that plays flac. No internet. No streaming. Just flac on a PC at the listening position > fiber > eR > RedNet D16 AES > Mutec MC3+ USB (WC & Re-clocker) > DAC.

 

Naturally this post is about what goes on in the computer.

 

Having re-established my PC after a bit of a nightmare with (as it turned out) a Network Adapter driver (see recent Optical Networks thread), I was feeling pretty pleased with SQ – thinking it was better than ever – and I was in a happy place. My spiking/dropouts problem was resolved – as proven in Dante Controller demonstrating stability well > 24 hrs.

 

 

Incidentally the Graph is even more "pure" with Flow Control On –

 

but SQ is definitely better with FC Disabled. Maybe the eR translating Gigabit to 100 Mbps benefits, as it were, from FC.

 

What about  the AXF-107 then?

 

mobo is Asus ProArt B550-CREATOR.

CPU is Ryzen 5600G.

RAM is CMK64GX4M2D3600C18.

 

mobo config:

CPU: 2 x (PCIe 4.0 x 16) [16 or dual 8] + M.2_1 PCIe 4.0

Chipset: 1 x (PCIe 3.0 x 16), (2 x PCIe 3.0) + M.2_2 PCIe 3.0

In first PCIe slot is Optane 900p with o/s.

flac Library is on WD Black SN850X 4Tb in M.2_1.

Network Adapter (single SFP) is Startech PEX1000SFP2 in PCIe_2.

 

The AXF-107 takes up no lanes – and does not appear in DevMan. No installation. It just sits in whichever PCIe slot you choose. It has LEDs which you can manipulate via DIP switches. Fortunately you can turn off the LEDs.

 

The AXF-107 has a "long" side and a "short" side.

 

First I tried the AXF-107 long side down in PCIe_3. Definitely meh. Sound was dry and robbed of body and power. Turned off the PC and went to bed.

 

This morning I flipped the AXF-107 (bracket moves to other end) so that it went short side down in one of the junior PCIe slots – more specifically – the one situated exactly between 900p and Network Adapter. Sits over the CMOS battery.

Different story. Jaw-dropping detail. Instruments and sound effects jumping out of the speakers. Voices laden with timbre. Even tighter and more powerful bass – which I experience somatically at my listening position thanks mostly to cherished Snell Type A III.

 

imo I could ABX this effect. SQ delta easily ranks alongside hardware upgrades.

 

Your mmv – of course. But, like Linus Gabriel Sebastian on this topic (albeit with different appreciations), I wasn't born yesterday – and this surprised me quite a lot. For the money – I would recommend AXF-107 as a potential SQ/$-£- € record breaker.

 

 

 

 

  

[MarcelNL]

first question in my mind is; what do you feed your computer with?

 

In a 'standard PC' setup with SMPS type PSU  that filter undeniably has an impact, I tried it when I finishd the unregulated LPSU feeding the Taiko ATX with bypasses and it did nothing (well, it did something that i did not particularly like)

[Iving]

1 minute ago, MarcelNL said:

first question in my mind is; what do you feed your computer with?

 

In a 'standard PC' setup with SMPS type PSU  that filter undeniably has an impact,

 

Paul Hynes SR7T Serial # 1

19V to HDPLEX DC-ATX 800W / Mundorf Silver-Gold loom

12V direct to CPU / Mundorf Silver-Gold

 

3 minutes ago, MarcelNL said:

I tried it when I finishd the unregulated LPSU feeding the Taiko ATX with bypasses and it did nothing (well, it did something that i did not particularly like)

 

Yes - my first trial I didn't like at all.

 

A different orientation / location on the mobo ... something else. Did you experiment much?

 

atm I conclude there is more scope than generally realised for cleaning up "dirty" mobo power.

[Exocer]

.

[Exocer]

8 minutes ago, MarcelNL said:

first question in my mind is; what do you feed your computer with?

 

In a 'standard PC' setup with SMPS type PSU  that filter undeniably has an impact, I tried it when I finishd the unregulated LPSU feeding the Taiko ATX with bypasses and it did nothing (well, it did something that i did not particularly like)

same for me, i tried it and did not like what it did with the ULPS V3 / Taiko dc-ATX.

 

Cheap enough to explore , for those curious!

[Iving]

3 minutes ago, Exocer said:

Same for me. I tried it and did not like it with the Taiko dc-ATX / ULPS v3 combo.

 

Perhaps my success has to do with AXF-107 in situ amongst 3 x CPU-direct parts/functions performing a single local playback role.

[Exocer]

2 minutes ago, Iving said:

 

Perhaps my success has to do with AXF-107 in situ amongst 3 x CPU-direct parts/functions performing a single local playback role.

Could be the case. Either way if I can local it I’ll try it again. I have many more reee slots now.

[MarcelNL]

I did not experiment much with it as i found other things to have much more impact on SQ than the location on the MB...the majority of caps on that thing are electrolytics, so turning them 180' seems counter intuitive to me...

My system is local playback and streaming but without upsampling/filtering too.

 

Have to add that I'm not a Mundorf S/G cap or S/G wire fan either.

[Iving]

8 minutes ago, MarcelNL said:

location on the MB

 

2 different slots were like -ve and then +ve otoh. without trying different possibilities I could easily have missed a SQ jump.

 

8 minutes ago, MarcelNL said:

turning them 180'

 

I had to flip AXF-107 to use in a PCIe x 4 slot (short side down) vs a PCIe x 16 one (long side down). Maybe not so much to do with turning caps 180 as reducing contact points in PCIe slot ... and 2nd slot being located as was first suggested to me - adjacent instead of remote from other Cards

 

8 minutes ago, MarcelNL said:

Have to add that I'm not a Mundorf S/G cap or S/G wire fan either.

 

:-) well I doubt that the wire has much to do with it - what have you moved on to?

[Exocer]

54 minutes ago, Exocer said:

Could be the case. Either way if I can local it I’ll try it again. I have many more reee slots now.

Locate it*
 

Free slots *